Automated Underwater Image Restoration Via Denoised Deconvolution

Description (excerpt)

CROSS-REFERENCE This application is a continuation of U.S. patent application Ser. No. 13/470,981, filed on May 14, 2012, entitled Automated Underwater Image Restoration Via Denoised Deconvolution, which is a divisional of U.S. patent application Ser. No. 12/418,683, filed on Apr. 6, 2009, entitled Automated Underwater Image Restoration Via Denoised Deconvolution, which application claims the benefit of priority based on U.S. Provisional Patent Application No. 61/043,302 filed on Apr. 8, 2008, the entirety of all the above-referenced documents is hereby incorporated by reference into the present application. TECHNICAL FIELD The present invention relates to image processing and image restoration, particularly relating to images of objects in and through a scattering medium such as images of objects under water. BACKGROUND The quality of images taken under water is vital to many military and civilian applications involving mine detection, diver visibility, and search and rescue. The ability to obtain better images at greater distances has often been a central goal of underwater imaging projects. Unlike in the atmosphere, where visibility can be on the order of miles, the visual range in the underwater environment is rather limited, at best on the order of tens of meters, even in the clearest waters. This is the result of the combined attenuation effects from both absorption, i.e., photons being absorbed into water molecules, phytoplankton cells, and detritus, and scattering, i.e., photons being bounced away from the original path into different traveling directions. It is mostly the effects of scattering by water and particulates that make the water look dirty or less transparent, resulting in a blurred image seen by human eyes and recorded by cameras. Image quality representation is an interesting and important research subject in digital image processing, especially with the rapid expansion of digital cameras, scanners, and printers into the everyday life of most households in recent years. Such devices would be of little use if they did not provide an acceptable representation of the subject of the image that was suitable for its intended purposes. The ability to objectively differentiate qualities amongst different images is critical in digital image processing, both for post-processing restoration of degraded imageries and in real-time imaging enhancement. A widely used criterion to evaluate image quality is the sharpness of the image, which represents the ability to reproduce details of subjects in the image. This directly affects the image's resolution, which is often expressed in terms of smallest pixels or the inverse of the highest spatial or angular frequency of the imaging system. Another related quality measure is the contrast of an image, and is usually determined by the difference between lighter and darker areas, normalized by the averaged brightness of both areas. See W. Hou et al., “Why does the Secchi disk disappear? An imaging perspective,” Opt. Exp. 15, 2791-2802 (2007) and H. H. Barrett et al., Foundations of image science (Wiley-Interscience, Hoboken, N.J., 2004). The most significant contributor to image blur is scattering, especially multiple scattering, where the path of a photon changes several times before reaching the receiver. The reduction in image quality due to scattering is two fold. Firstly, the un-scattered direct beam which contributes to the sharp part of the image is correspondingly reduced. Secondly, the scattered photons help to brighten the previously darkened area thus reduce contrast. Adding absorption on top of scattering, the reduction in signal can be so great that the electronic noise of the system becomes a factor, further complicating the issue. The amount of blurring in an image can be described by how much blur a point-source will introduce over the imaging range. This property is the point-spread function (PSF) of the imaging system. The Fourier transform of the PSF is known as the optical transfer function (OTF), generally for incoherent imaging without considerations of phase information, the magnitude of OTF, referred to as the modulation transfer function (MTF), often is used. The OTF (or MTF) describes the frequency response of signals over transmission range, or how fast the details of an image degrade in a given environment. To compensate for blur and improve imagery effectively, it is critical to incorporate knowledge of the optical properties of the water to better model the degradation process. Studies have been done regarding image degradation through the atmosphere transmission. However, unlike the underwater environments, degradations of the image quality by the atmosphere are most dominantly caused by turbulence under optimal conditions, although scattering by particles and aerosols also play a minor role. Better restoration in astronomy or reconnaissance applications can be obtained with knowledge of the modulation transfer functions.